Nine AI-Resilient In-class Assessment Activities
Generative AI has created a situation where many feel that more assessment must be done in-class, but not all courses lend themselves to oral exams or presentations. Here are nine activities you can use in courses of many sizes to gather information about what your students are learning. The real-time nature of these activities allows no time for outsourcing responses.
Don’t see something exactly right for you?
These materials were created in collaboration with Claude AI, starting with the prompt:
You’re an expert at designing creative in-class assessment activities for the university classroom. I am a faculty developer at an R1 university and I need nine concrete examples of activities that a faculty member can use for in-class assessments, even for very large classes. NOTE: I need activities that are AI-proof and scalable. This means no oral exams or presentations.
If any of these activities seem close to useful but not-quite-right for your context, try pasting from this page into a Claude chat and ask for suggestions that could bring it closer to “just right” for the details of your needs.
In this guide:
- 1. Cold-Text Annotation & Analysis
- 2. Concept Map from Memory
- 3. Error Detection & Correction
- 4. Live Data / Scenario Interpretation
- 5. Structured Argument Ranking
- 6. Muddiest Point + Peer Explanation
- 7. Course Concept Applied to a Local/Campus Context
- 8. Before/After Knowledge Comparison
- 9. The Discipline-Specific Sketchnote
The Activities
1. Cold-Text Annotation & Analysis
What it is: Distribute a passage, dataset, or image on paper that students have not seen before at the start of class. Students annotate it directly and respond to 3–5 targeted questions written below.
How it works:
- Instructor selects a never-before-shared text or image (a news excerpt, a graph, a legal clause, a poem)
- Students annotate for 10–15 minutes, then answer short-answer questions on the same sheet
- Collected at the end of class
Why it’s scalable: Same paper distributed to all; fast to grade with a rubric focused on quality of reasoning, not recall.
Grading tip: Use a 3-point holistic rubric per question (missing → partial → complete reasoning).
Some Discipline-Specific Examples
2. Concept Map from Memory
What it is: Students are given a blanxk sheet with a central term or concept and must build a concept map connecting at least 8–10 course terms — from memory, in class.
How it works:
- Provide a single-page template with the central concept pre-printed
- Students have 15–20 minutes to construct the map using labeled links between nodes
- Maps are collected and graded on accuracy of connections and the quality of linking language
Why it’s scalable: No prep beyond printing. Maps reveal misconceptions quickly and are easy to grade holistically.
Grading tip: Develop an “anchor map” with must-have connections. Award points for each valid connection (e.g., 1 pt each, up to 15 pts total).
Some Discipline-Specific Examples
3. Error Detection & Correction
What it is: Students receive a worked problem, short essay, experimental design, or argument that contains deliberate errors and must identify, explain, and correct them.
How it works:
- Embed 3–5 errors of varying subtlety (conceptual, methodological, factual, logical)
- Students write: (a) what the error is, (b) why it’s wrong, and (c) the correction
- Collected at end of class
Why it’s scalable: Forces higher-order thinking, not recall. Extremely fast to write and reuse across semesters with minor modifications.
Grading tip: Assign equal point values per error. Require explanation — circling alone earns no credit.
pts total).
Some Discipline-Specific Examples
4. Live Data / Scenario Interpretation
What it is: Instructor presents a data table, graph, image, or case scenario for the first time in class and students interpret it on paper in a specified time window.
How it works:
- Distribute or project a novel artifact (one they’ve never seen)
- Pose 2–4 structured questions: What pattern do you observe? What conclusion can you draw? What is a limitation of this data?
- Students write responses individually on printed paper or index cards
Why it’s scalable: The prompt never leaves the classroom. No answer key exists online. Easily adapted across disciplines (clinical data in nursing, a primary source in history, a balance sheet in business).
Grading tip: Grade on reasoning process, not a single “right” answer. A 2-point scale (reasoning present/absent) per question moves grading fast.
Some Discipline-Specific Examples
5. Structured Argument Ranking
What it is: Students receive 4–5 pre-written arguments or solutions and must rank them from strongest to weakest, with a written justification for each ranking decision.
How it works:
- Provide a brief scenario or question at the top of the sheet
- List 4–5 plausible responses below (some strong, some flawed)
- Students rank them AND write 2–3 sentences justifying each rank position
Why it’s scalable: Even if students discussed the question beforehand, the specific set of options is novel and the justification is what’s graded — not the rank itself.
Grading tip: Don’t grade the ranking — grade the quality of the justification. A student who ranks well but explains poorly scores lower than one who ranks differently but reasons sharply.
Some Discipline-Specific Examples
6. Muddiest Point + Peer Explanation
What it is: A two-stage written activity where students first write the recent course material they find most difficult or confusing, then swap papers and write a substantive response to their peer’s confusion.
How it works:
- Stage 1 (5 min): Each student writes their single muddiest point from the unit — specific, not vague (“I don’t understand X because I thought Y but the reading says Z”)
- Stage 2 (8–10 min): Papers are swapped; student writes a teaching response to the confusion they received
- Both papers collected; both students graded
Why it’s scalable: Generates 100–200 graded items from one activity. Also gives the instructor a real-time diagnostic of class-wide confusion.
Grading tip: Grade Stage 1 on specificity; grade Stage 2 on accuracy and helpfulness of the explanation. Each stage ~5 pts.
Some Discipline-Specific Examples
7. Course Concept Applied to a Local/Campus Context
What it is: Students apply a course concept to a specific, local context — a campus policy, a local news story, a building on campus, an event that happened that week — that AI cannot know or fabricate accurately.
How it works:
- Instructor identifies a local/timely reference in advance (e.g., “The university recently announced X. Apply the concept of Y to analyze this decision.”)
- Students write a 1–2 paragraph analytical response on paper
- Collected at the end of class
Why it’s scalable: The local anchor makes it ungoogleable and un-AI-able. Works in every discipline. One well-designed prompt can serve a full class.
Grading tip: Build a rubric around: (1) accurate use of the concept, (2) quality of the local application, (3) depth of analysis. Ignore surface writing quality for speed.
Some Discipline-Specific Examples
8. Before/After Knowledge Comparison
What it is: Students complete a short written task before a lesson or class activity and then repeat (or respond to) a parallel task after — demonstrating learning that occurred in real time.
How it works:
- At the start of class: “In 3–4 sentences, explain [concept X] as you currently understand it.”
- Conduct the lesson/activity
- At the end of class: “Now revise your explanation. What did you change, add, or correct? Why?”
- Both entries collected and graded together
Why it’s scalable: The “after” response is only meaningful if the student was present and engaged. The revision itself is the evidence of learning.
Grading tip: Grade the quality and specificity of the revision, not the initial entry. Award points for articulating what changed and why — not just rewriting it better.
Some Discipline-Specific Examples
9. The Discipline-Specific Sketchnote
What it is: Students create a one-page visual summary — using both words and hand-drawn diagrams — of a concept, process, or argument introduced in that class session.
How it works:
- At the end of a lecture or discussion, provide 10–12 minutes and a blank half-sheet
- Students must include: a title, at least one diagram or visual, labeled components, and a written takeaway
- Collected at the end of class
Why it’s scalable: AI cannot produce handwritten, hand-drawn artifacts. Sketchnotes are personal, idiosyncratic, and verifiable as in-room work. Particularly effective in STEM (process diagrams), social sciences (concept relationships), and professional fields (workflows).
Grading tip: Use a 3-category rubric: Visual accuracy (is the diagram correct?), Completeness (are key elements labeled?), Takeaway quality (is the written insight meaningful?). Avoid grading artistic quality.